By Nila Kartika Wati 
In a dimly lit chamber at Imperial College in London during the late 1920s, a small group of ten individuals meticulously observed a split circle of light, no wider than a thumbnail held at arm’s length. One half of this illuminated disc emanated a single, pure wavelength of light, while the other comprised a mutable blend of red, green, and blue lamps. Their task was to manipulate the intensity of these primary colors until the two halves seamlessly merged into a uniform patch. Concurrently, seven more participants undertook an identical exercise at the National Physical Laboratory in Teddington. The numerical data meticulously generated by these 17 individuals forms the bedrock of every ICC profile on your computer today, a testament to a scientific endeavor that, almost a century later, continues to dictate how we perceive and reproduce color in the digital realm.
This chain of influence is remarkably short and unbroken. Contemporary color spaces such as sRGB, Adobe RGB (1998), Display P3, Rec. 2020, and ProPhoto RGB are all meticulously defined by their red, green, and blue primaries, pinpointed as xy coordinates on the venerable CIE 1931 chromaticity diagram. CIELAB, a widely used color space for perceptual uniformity, is itself a transform of CIE XYZ. The Delta E values, which countless photographers and calibration specialists rely upon, represent distances measured within the CIELAB space. At its core, ICC color management orchestrates all these processes through a profile connection space, typically built upon CIE XYZ or CIELAB, standardized under illuminant D50 and the foundational CIE 1931 2-degree standard colorimetric observer. Untangle any thread of this intricate system, and one invariably arrives back at those handful of British laboratory volunteers, diligently turning knobs in pursuit of a perfect match.
The Genesis of a Standard: Imperial College and NPL
The experimental setup employed by these pioneering scientists was a trichromatic colorimeter, designed for a task known as a bipartite field match. Observers faced a circular field bisected into two semicircles. One side displayed the "test light," a narrow band meticulously extracted from a spectrum. The opposing side presented an additive mixture of three primary lights, whose intensities were under the direct control of the observer. The objective was deceptively simple yet profoundly challenging: to make the visible seam between the two halves vanish, indicating a perfect color match.
W. David Wright led the Imperial College contingent, utilizing monochromatic primaries at precise wavelengths of 650, 530, and 460 nanometers. His team systematically stepped through the entire visible spectrum, from 400 to 700 nm, in 10 nm increments. In parallel, John Guild directed the efforts at the National Physical Laboratory. Guild’s team opted for broadband primaries, achieved by placing red, green, and blue gelatin filters in front of an opal-bulb lamp, and worked across the spectrum from 380 to 700 nm in finer 5 nm intervals. This involved two distinct laboratories, employing two different instruments, and two separate sets of primary lights, yet ultimately involving the same core challenge for 17 human observers. Wright’s seminal findings were published in his 1929 paper in the Transactions of the Optical Society, while Guild’s equally crucial work appeared in the Philosophical Transactions of the Royal Society A in 1931.
The profound agreement between these two independently derived datasets was striking, particularly given the relatively small sample size involved. This unexpected concordance provided the crucial impetus for an international body to take the bold step of standardization. At its eighth session, convened in Cambridge in 1931, the Commission Internationale de l’Éclairage (CIE) officially adopted the combined data as the "standard colorimetric observer," effectively laying the immutable foundation for the entire field of colorimetry.
A key challenge emerged during these experiments: certain spectral test lights, particularly highly saturated ones, could not be matched by any positive mixture of the three real primaries. The additive mixture consistently appeared too pale. The ingenious solution was to introduce a "negative" primary. This involved moving one of the primaries across the seam, adding it to the test light side, and then recording its contribution as a negative number. This methodological necessity is precisely why the CIE RGB color-matching function for red dips below zero across a significant portion of the blue-green region. It also explains why the gamut of any three-primary display is inherently a triangle, forever contained within the horseshoe-shaped spectral locus, which it can never fully encompass. The introduction of additional primaries, as seen in multiprimary LCDs that incorporate yellow and cyan channels, transforms the gamut into a convex polygon, still confined but reaching further into the visible spectrum.
Unpacking the Foundations: The 1931 Observer’s Limitations
The popular narrative often portrays the standard observer as a statistically robust average of human vision, a careful and comprehensive portrait of normal color perception. This widely held belief, however, overlooks critical nuances. The sample size of 17 volunteers, drawn exclusively from two English physics laboratories, was never a secret. These individuals were meticulously screened for normal color vision and were adults available for extended sessions in darkened rooms during the 1920s. Guild himself acknowledged in his paper that Wright’s ten observers exhibited a closer agreement with the seven from the National Physical Laboratory than one might statistically expect from groups of their size, a statement that hints at a degree of good fortune as much as biological consistency.
Yet, the sample size is not the most profound issue. The CIE 1931 color-matching functions were not, in fact, measured directly in their entirety. Both Wright and Guild produced relative chromaticity data, which accurately defined the shape of the color triangle but did not fully establish the perceived brightness associated with it. To transform this into a complete set of color-matching functions, the international committee introduced a crucial constraint: the y-bar curve, representing the luminous efficiency function, was forcibly equated to the 1924 CIE photopic luminous efficiency function. This 1924 curve was the then-standard for how bright a watt of light appeared at each wavelength. Consequently, the 1931 observer is, in essence, a reconstruction, meticulously assembled by fitting one laboratory’s chromaticity data to a brightness curve derived from a different era and different experimental conditions. The Colour and Vision Research Laboratory at University College London has openly described both the borrowed brightness curve and the resulting reconstruction as "questionable," highlighting a fundamental compromise at the heart of the standard.
The 1924 brightness curve itself had a complex provenance. It was not derived from the 17 observers but rather compiled from luminosity measurements taken by various groups using a multitude of different methodologies. Thus, while the headline figure of 17 people accurately reflects the origin of the chromaticity half of the standard observer, it significantly understates the intricate and somewhat fragmented origins of the brightness component. Seventeen people established the fundamental shape of color perception, but a committee’s compromise, drawing from separate and diverse experiments, determined the critical brightness weighting that permeates the entire system.
Why the Blues Were Always the Weak Spot
This borrowed brightness curve is the direct progenitor of the most well-documented defect within the entire CIE 1931 system. The 1924 function demonstrably underestimates human sensitivity in the violet region of the spectrum, with the magnitude of this error approaching an order of magnitude at the shortest wavelengths. Because the y-bar curve was rigidly defined to equal this flawed function, the error was intrinsically woven into the luminance channel of CIE XYZ, and subsequently propagated into CIELAB lightness, into Delta E calculations, and ultimately into every color profile that relies upon these standards.
The color science community has been acutely aware of this significant deficiency since at least 1951, when Deane Judd published a revised luminous efficiency function, proposing corrections specifically for wavelengths below 460 nm. J. J. Vos further refined Judd’s version in 1978. Despite these advancements, neither replacement was adopted as the general standard for imaging. One reason for this reluctance was that Judd’s correction introduced its own artifact: the modified curve implied a standard observer with an unrealistically high macular pigment density, effectively trading one distortion for a slightly smaller, albeit different, one.
For the vast majority of photographic applications, the practical consequences of this blue-end error remain modest, primarily because most captured light is broadband, and the error resides at the extreme edge of the visible spectrum. However, this modesty vanishes when dealing with narrow-band light sources. Deep blue and violet LEDs, advanced laser projection systems, intensely saturated blue stage lighting, and the blue primary of wide-gamut displays all operate perilously close to the spectral region where the 1931 functions are demonstrably least reliable. If a professional has ever measured a deep blue tone and received a numerical value that felt intuitively incorrect compared to their visual perception, it is highly probable that their intuition was, in fact, accurate.
Beyond the Fovea: The 10-Degree Observer of 1964
The original CIE 1931 standard was predicated on a 2-degree viewing field, which deliberately restricts the stimulus to the central fovea of the eye. This region is characterized by the highest density of cones and a scarcity of rods, the photoreceptors responsible for low-light vision. At a typical monitor viewing distance of approximately 24 inches, a 2-degree field subtends a patch roughly 0.8 inches across. Crucially, almost nothing that a professional evaluates on a screen in real-world scenarios is confined to such a minuscule area. The skin tones in a full-page portrait, a expansive sky, a wall of product packaging, or a print held at a standard reading distance—all constitute large fields. These large fields impinge upon the retina with varying amounts of macular pigment, a yellow screening pigment that is densest over the central fovea and progressively thins out beyond it. Consequently, a wide field reaches the cones through less of this pigment, leading to different color matching characteristics.
Recognizing this fundamental limitation, the CIE introduced a supplementary observer in 1964, specifically designed for 10-degree fields. This later standard was built upon a significantly more robust and directly measured dataset. While rods are present in this wider retinal field, their influence was deliberately excluded from the standard’s description. The 1964 observer is specified for high photopic (daylight) levels and for spectral conditions where rod participation is not anticipated, with experimenters taking pains to suppress any rod influence in the color matching process itself. W. S. Stiles and J. M. Burch, working at the National Physical Laboratory, meticulously measured 49 observers across the spectrum from 392.2 to 714.3 nm, using primaries near 645, 526, and 444 nm. Their color-matching functions were derived directly from these measurements, a distinct improvement over the reconstructed nature of the 1931 standard. Their comprehensive final report was published in 1959, the same year N. I. Speranskaya published similar matches from 27 observers. The 1964 observer primarily relies on the Stiles and Burch data, supplemented by Speranskaya’s findings, with the CIE performing adjustments and smoothing before final publication. The resulting recommendation was straightforward: use the 10-degree observer when the color-matching field exceeds approximately 4 degrees.
The Paradox of Precision: Industry Inertia and Practical Consequences
Despite the clear scientific advancements embodied by the 1964 10-degree observer and subsequent, even more refined models (such as the cone-fundamental-based functions in CIE 170-1 from 2006 and CIE 170-2 from 2015), practical industry adoption has been minimal. Your camera profiles, display profiles, printer profiles, soft proofing tools, and Delta E readouts almost universally continue to rely on the 2-degree observer from 1931. The question of whether this is the "wrong instrument" for evaluating photographs is more complex than a simple comparison of viewing angles might suggest. The 4-degree rule applies to a specific matching field, not necessarily the entire width of a picture. Research with complex images suggests that effective field size for such scenes can appear smaller than their physical screen area. While a mismatch exists, it’s not a tidy 30-degrees-against-2 arithmetic. Writing in LD+A, the journal of the Illuminating Engineering Society, Jess Baker, Tony Esposito, and Jason Livingston succinctly attribute the 1931 observer’s enduring survival to the immense inertia that must be overcome before any established metric is changed. The global cost and logistical complexity of retooling hardware, software, educational curricula, and professional workflows represent a formidable barrier to adopting a more accurate, albeit newer, standard.
The Human Element: Observer Metamerism and Individual Perception
None of this inherently renders the standard observer "wrong." Rather, it highlights that it represents a single, fixed answer to a question that, in reality, encompasses a distribution of answers. Human color matching varies significantly even among individuals with entirely normal color vision. These variations are driven by factors such as macular pigment density, the optical density of cone photopigments, a common genetic polymorphism in the long-wavelength cone, and the natural yellowing of the eye’s lens with age. Consequently, two individuals can look at the same patch on the same calibrated monitor, perceive genuinely different colors, and both be accurately describing their unique retinal experience.
Color scientists term this phenomenon "observer metamerism," and its effects become more pronounced as display primaries become spectrally narrower. A modern display or projector whose red, green, and blue outputs are spectrally narrow spikes will indeed produce a wide gamut, but it will also amplify the perceptual differences between two observers. A subtle shift in an individual’s cone spectral sensitivities means they will catch a different amount of a narrow spectral spike, leading to disparate color perceptions. Modeling work using measured individual cone functions has estimated average observer metamerism indices for laser projection systems in the range of approximately one to two units of color difference, with maxima reaching around three. Near-neutral bright tones are often among the areas where individual observers disagree most. While these numbers might not appear enormous in isolation, they are comfortably larger than the tight tolerances professionals frequently debate in color-critical review environments.
The very instruments used for color measurement inherit these underlying assumptions. A display colorimeter, such as the Calibrite Display Plus HL, reports tristimulus values by approximating the 1931 color-matching functions with internal filters. This is precisely why calibration software typically ships with specific correction data tailored to different display technologies. The puck is not directly measuring "color" as a universal truth; it is measuring light and then interpreting that light through the lens of what 17 English volunteers would have reported a century ago.
Navigating Color Disputes in the Digital Age
A numerical value within a color-managed pipeline is fundamentally a prediction about a hypothetical observer, not an absolute description of the person standing behind you. Delta E, in its essence, answers the question of whether two stimuli would have matched for the 1931 average observer. It does not, and cannot, answer whether they match for your specific client, nor can any calibration make it do so. As the LD+A authors poignantly articulate, "It is simply not possible to create a color match for all people simultaneously, even if all viewers have normal color vision."
This has concrete and profound consequences for how professionals conduct color reviews. When a client expresses that skin tones have acquired an undesirable green cast, and your calibrated numbers assert otherwise, responding with "the numbers win" is not only impolite but, strictly speaking, scientifically unjustified. What the numbers unequivocally establish is that your file will behave predictably and consistently as it transitions between various devices—which is the fundamental objective color management was originally invented to achieve. Appearance disputes, however, are best resolved by bringing the disputing parties into a shared retinal experience: in front of the same display, under identical room lighting, at the same viewing distance, and observing the same size patch. By shifting the argument from abstract numerical values to a common, direct visual experience, the vast majority of such disagreements often evaporate.
It is crucial to maintain a clear distinction between the technical conversation and the aesthetic conversation, as the former often possesses a "right answer" while the latter is inherently subjective. When possible, professionals should lean on relationships rather than absolute values. A retoucher who has honed their craft over years of color grading portraits is keenly observing how a skin tone harmonizes with the background and wardrobe. These intricate relational aspects of color often survive observer differences far more robustly than any single absolute coordinate.
The widely used ProPhoto RGB color space makes the abstract nature of color perception impossible to ignore. Two of its three primaries lie entirely outside the spectral locus, meaning they correspond to no physically achievable light. A commonly cited figure suggests that approximately 13% of the color codes within the ProPhoto space represent colors no human being can actually perceive. These seemingly "unphysical" coordinates exist because the horseshoe-shaped map, derived from those 17 individuals’ knob settings in 1931, serves as the ultimate boundary. Kodak’s engineers deliberately placed these "corners" outside the visible spectrum to encompass the full range of real-world surface colors that a photographic image must be capable of holding. Yet, even with this expansive approach, the ProPhoto triangle still misses a portion of the horseshoe: highly saturated blue-greens near 500 nm fall outside its bounds. Every time a ProPhoto file is opened, one is working within a container whose edges were ultimately fixed by the fundamental limits of what a small group of volunteers in two English laboratories could match with three lamps almost a century ago.
The profound impact of these early experiments continues to resonate through every pixel and print, a silent testament to the foundational work that underpins our modern visual world. While the tools and understanding of color science have evolved dramatically, the specter of those 17 volunteers in darkened rooms remains, a powerful reminder of the human element at the heart of even the most sophisticated technological systems.
